Loading blog posts...
Loading blog posts...
Loading...

Programming language changes for the week of 20 November 2026 make invalid states, risky dependencies and unsupported deployment assumptions harder to ignore. Those boundaries should guide your next upgrade decision.
Quick summary
Rust 1.100, released on November 12, combines new type-system capabilities with Cargo controls. For application teams, the dependency publication cooldown deserves attention before the allocator work.
toml#.cargo/config.toml [registry] global-min-publish-age = "3 days"
This setting prevents Cargo from selecting newly published dependency versions until they have aged past the configured threshold. It gives teams a pause before accepting fresh registry releases. It doesn’t scan packages for security issues or guarantee that older packages are safe.
cargo install ignores the cooldown.[registry] in .cargo/config.toml.Rust now permits Result<T,!>, where the never type, !, represents a value that cannot exist; Infallible also becomes an alias of !.
! in editions 2015, 2018 and 2021.Allocator trait is stabilized for custom allocation with Vec and Box.Don’t introduce custom allocation merely because the API is stable. Confirm that allocation policy is a requirement and that every collection and pointer type in your design has the support you need.
Code that depends on type inference can be affected by fallback changes without an edition migration. Run the new compiler against your existing workspace before combining the upgrade with dependency updates.
C# 15 shipped with .NET 11 on November 10, adding unions and compiler-checked exhaustive switch expressions. It’s the default language version for projects targeting net11.0.
A product lookup might return either a product or an explicit not-found result. A union declares those alternatives directly, removing the need for callers to infer them from a nullable value or an exception convention.
public union ProductLookup(Product, ProductNotFound).Use unions for expected alternatives. A missing product is a normal domain outcome; a broken database connection is a different category of failure. There’s no reason to convert every exception into a return value.
Adding another declared outcome gives the compiler a way to identify switch expressions that need attention. Framework selection alone can’t solve this domain-handling problem; the meta-framework comparison covers the application-layer choices.

Kotlin 2.5.0-RC, released on November 11, enables its unused return value checker by default. Its experimental full value classes need a higher adoption threshold.
-Xfull-value-classes.17 raises an error.A discarded return value can hide a missing step: your code calls a transformation, ignores its result and continues using the original value. Stronger checking catches that mistake closer to the edit that introduced it. Inspect the diagnostics before treating every warning as a defect.
Test the checker before adopting experimental representation changes. Put the release candidate on a separate CI lane and inspect the findings.
Check the CI image and build-tool configuration for the JDK running the compiler. Changing a project’s target doesn’t establish which JDK runs its compiler.
Dart 3.14 adds automated SDK migration commands and standalone WebAssembly output.
Preview the migration first:
bashdart migrate --dry-run
Inspect the proposed changes and keep dependency upgrades out of the same patch where possible, so a failed test points to fewer possible causes.
Apply the migration on a clean branch once you’ve reviewed the changes:
bashdart migrate --apply --target-sdk=3.14.0
The apply command performs preparatory fixes and raises package SDK constraints for the target release. Review the generated diff and run tests.
The deployment change needs a separate compatibility check:
dart compile wasm --standalone produces standalone Wasm output.128-bit Wasm SIMD instructions.Removing the JavaScript host requirement changes where the binary can run. Your application still needs compatible runtime facilities.
Test a candidate service or command-line artifact inside the intended runtime, including its filesystem access, inputs and deployment packaging.

The November 11 Java announcement says primitive patterns will be unavailable in JDK 28, even with --enable-preview. The designers want numeric conversion rules to account for Valhalla’s evolving numeric types before committing further.
28.Withdrawing the feature is preferable to a hurried commitment to rules that become expensive to preserve. Numeric matching needs predictable conversion behavior, especially when a conversion changes which values are representable.
Identify code and teaching material that depends on the preview. Don’t plan a JDK upgrade around keeping that code working through a preview flag.
The November 15 benchLANG comparison reports closely grouped results for C, Zig and Rust across its workload set.
| Language | Reported time | Measurement context |
|---|---|---|
| C | 4.35 seconds | Seven workloads; average of two runs |
| Zig | 4.59 seconds | Seven workloads; average of two runs |
| Rust | 4.73 seconds | Seven workloads; average of two runs |
The comparison uses Nix-pinned toolchains, includes JIT warm-up and permits compile-time constant folding. It measures particular programs under those rules.
Constant folding is legitimate compiler work. If a compiler calculates an answer before execution, though, the result may tell you little about an application whose inputs arrive from users or a network.
Two other November sources have their own measurement conditions:
1.83x improvement on its typed-array benchmark after vectorizing eligible loops.--stdout-buffering option enables a 64 KiB buffer for non-terminal standard output.The ScriptC result is useful if your workload contains the loop shapes that benefit. It says nothing about unrelated code, and output buffering is a separate optimization with different effects.
Choose a language for its constraints, then benchmark the implementation. Include startup for a short-lived command; test steady-state behavior separately for a long-running service where needed. Validate correctness before measuring either.

cargo install, which bypasses it.1.100 before updating dependencies; investigate inference-related failures separately from dependency changes.2.5.0-RC in a non-blocking CI job using JDK 17 or newer, and review each new unused-return diagnostic.Judge your next language upgrade by which assumptions it makes checkable and which compatibility obligations remain yours. Feature counts won’t tell you whether it improves your engineering process.